SEARCH

SEARCH BY CITATION

Keywords:

  • behaviour;
  • hunting spider;
  • intraguild interactions;
  • kairomone;
  • non-consumptive effects;
  • predation;
  • silk;
  • web builder

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Predators can induce changes in prey phenotype such as dispersal, activity and foraging rate. Such trait-mediated effects (TMEs) can strongly affect prey populations and generate trophic cascades, rivaling the importance of predation in communities. However, the relevance of TME on intraguild interactions has rarely been addressed. Ants and spiders are widespread generalist predators in terrestrial habitats. Ants influence arthropod assemblages and disrupt top-down effects of spiders on herbivores by killing spiders and/or by inducing spider emigration. Here, we examined whether ants induce dispersal behaviour in spiders. We tested the effect of chemical cues of two ant species (Lasius niger, Formica clara) on the walking activity and the propensity for silk-based dispersal of spiders. Silk-based dispersal of the web-builder Phylloneta impressa increased by 80% with exposure to Lasius cues, whereas dispersal of the hunting spider Xysticus more than doubled when confronted with cues of both Lasius and Formica. In addition, Xysticus individuals showed a marked increase in walking activity when exposed to Formica but not Lasius cues. Our results show for the first time that perceived predation risk influences spider dispersal. The strong effect of ant chemical cues on spider dispersal demonstrates that TMEs contribute to the impact of ants on arthropod communities.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Predators reduce prey abundance through consumption. A major endeavour of community ecology is studying these density-mediated effects (DMEs) of predators to quantify their impact in community dynamics, food web structure and biological control (Murdoch, Briggs & Nisbet, 2003; Morin, 2011). However, predators also influence their prey in a non-trophic way: because of the selective pressure to avoid predation, the mere presence of a predator can induce changes in prey traits to minimize predation risk, hence known as trait-mediated effects (TMEs; Lima, 1998; Werner & Peacor, 2003). These predator-induced phenotypic changes concern a great diversity of traits related to behaviour (e.g. activity level, dispersal), physiology (hormonal levels) and morphology (defensive structures), which have associated costs that prey balance to the risk of predation (Lima & Dill, 1990; Lind & Cresswell, 2005). Research on TME, at the intersection between behaviour and ecology, has recently shown TMEs to be major drivers of trophic cascades and to influence different aspects of ecosystem functioning, sometimes exceeding the importance of DME in communities (Schmitz, Krivan & Ovadia, 2004; Preisser, Bolnick & Benard, 2005; Schmitz, Hawlena & Trussell, 2010).

TME can even propagate beyond community boundaries when predators change prey migration rates (Diehl et al., 2000; Cronin, Haynes & Dillemuth, 2004). As a consequence, prey populations can be reduced by a combination of predator-induced mortality (DME) and dispersal (TME). So far, empirical evidence for the impact of TME on dispersal basically comes from aquatic systems, while terrestrial communities remain understudied (Orrock et al., 2008, 2010).

Ants and spiders are main components of the arthropod predator guild in terrestrial habitats. Ants play multiple important roles in communities as hemipteran mutualists, predators and ecosystem engineers (Stadler & Dixon, 2005; Sanders et al., 2011). Experimental field studies in different settings show that ants have a strong influence on arthropod assemblages by reducing the densities of insects and spiders and disrupting top-down effects of spiders on herbivores (Vandermeer et al., 2002; Sanders et al., 2011; Piñol, Espadaler & Cañellas, 2012a). Field experiments do not usually permit identifying the causes underlying the patterns revealed by the manipulations, thus, the mechanisms by which ants drive these changes (e.g. DME, TME, competition) are largely unknown. Our understanding of the processes that underlie intraguild interactions is particularly important owing to their prevalence and far-reaching impacts in food webs, like the emergence of non-additive effects and the changes in the strength of trophic cascades (Arim & Marquet, 2004; Vance-Chalcraft & Soluk, 2005). However, the relevance of TME on intraguild interactions has rarely been addressed (but see Schmidt-Entling & Siegenthaler, 2009).

TMEs commonly emerge when prey detect predator semiochemicals (Ferrari, Wisenden & Chivers, 2010). Ant semiochemicals are complex blends that are central to intraspecific communication, like nestmate recognition, cooperation, navigation and food recruitment (Lenoir et al., 2001), as well as to parasitic and mutualistic relationships (Lang & Menzel, 2011). The effects of semiochemicals deposited by patrolling ants can extend beyond these particular interactions and become cues that induce TME, but so far, empirical examples are very scarce and the identity of such semiochemicals is unknown (Morgan, 2009; Van Mele et al., 2009).

In the present work we investigate the hypothesis that TMEs are one of the mechanisms underlying the strong intraguild interference that ants exert on spiders. We focused on the effect of ants on spider activity and dispersal behaviour. Spiders are ideal model organisms to study dispersal behaviour because they have a unique strategy based on silk that can easily be quantified in the laboratory. To disperse, a spider climbs to an elevated point and stands on tiptoe, a stereotypical display that consists in stretching the legs and raising the tip of the abdomen to release silk into the air. The spider uses this silk as bridge between vegetation to travel up to several meters, a dispersal mode called rappelling. Many spiders are also able to disperse by ballooning, whereby they make a short silk thread to become airborne, then use it as a drag to disperse long distances based on wind currents. In addition, spiders select between microsites or even move between habitats by walking (Bell et al., 2005). In our study we tested the existence of TME of ants on spiders by assessing (1) whether cues from two ant species increase the propensity of juvenile cobweb and crab spiders to disperse using silk and (2) whether ant cues change the walking activity of crab spiders.

Material and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Rearing of individuals and experimental design

Between June and July 2012 we sampled arthropods from a grassland site south of the city of Landau (Rhineland-Palatinate, Germany). We took ant nests of Lasius niger Linnaeus and Formica clara (Seifert) because we commonly observed workers of both species patrolling on the ground and among the vegetation. We selected two common spider species that have a sedentary lifestyle, as defined by their hunting mode (Foelix, 2010): the cobweb spider Phylloneta impressa (Koch) (Theridiidae), a web-building species that makes tangle webs with its shelter to capture prey, and the crab spider Xysticus Koch (Thomisidae), a spider with a sit-and-wait capture strategy that does not build webs. We collected only small-sized juvenile spiders because TMEs are stronger when body mass predator:prey ratios are high (Preisser & Orrock, 2012; Binz et al., 2014) and because large, adult spiders can prey on ants (pers. obs.; Foelix, 2010). To obtain juvenile P. impressa we collected adult females with egg sacs that we found in Eryngium campestre Linnaeus (Apiaceae). In that period of the year, adult Xysticus females were rare, so we had to collect Xysticus juveniles by sweep netting among the grass. Because juveniles lack genital features that allow for species identification, our sample of juveniles contained a mixture of Xysticus species. Based on a small sample of field-collected adult females and of juveniles we reared into adulthood, these species were X. cristatus (Clerck) and X. luctuosus (Blackwall).

In the laboratory we kept ant nests in plastic terraria lined with plaster of Paris to maintain humidity. We fed them regularly with house crickets, Acheta domesticus, and honey. We kept all spiders individually in small Petri dishes (diameter: 3.5 cm), inside a climate cabinet (25°C, RH = 75%, L:D = 16:8). When egg sacs of female cobweb spiders hatched, we left the spiderlings with their mother for 15 days because females of this species feed their offspring by regurgitation for a few days (Kullmann, 1969). Then we transferred the young into individual Petri dishes. We fed the spiders with fruit flies Drosophila hydei, ad libitum twice per week. Just before the experiments, we measured prosoma width to the nearest 0.1 mm and randomly assigned spiders to treatments ensuring that the average size of individuals was equal across treatments (Xysticus: 0.81 ± 0.017 mm, P. impressa: 0.46 ± 0.0048 mm). We also equally distributed P. impressa juveniles of different broods between treatments to control for potential family effects on dispersal propensity.

Dispersal experiments

To test the effect of Formica and Lasius cues on spider dispersal propensity, we conducted separate experiments on the crab spider and on the cobweb spider. We built dispersal arenas by filling small Petri dishes (diameter: 3.5 cm) with plaster of Paris and inserting a wooden stick into them (length: 7 cm). We established treatments by enclosing six Lasius (Lasius cue experiment) or three Formica ants (Formica cue experiment) in the dispersal arenas with fluon-coated tops to make the ants remain on the bottom so that they would deposit cues while walking on the plaster of Paris. We used a pooter to introduce the desired number of ants into the arenas and enclosed them right away to prevent the ants from escaping. We enclosed control arenas with no ants inside and we kept all dispersal arenas inside a climate cabinet for 24 hours (25°C, RH = 75%, L:D = 16:8). Immediately before the dispersal experiments, we opened the arenas and removed the ants from the treatment ones so that only cues would be left on the surface. Because ants readily walked away from the arenas, we did not need to handle them.

We placed the spiders individually on dispersal arenas and we tested two individuals simultaneously, randomly assigned to either a cue or a control arena. We arranged the two dispersal arenas sideways, separated by a carton wall to avoid potential volatile cues from reaching the control arena. We put each arena in a water bath to prevent spiders from walking away. In front of the arenas, we placed a fan that produced an upward air current at an average speed of 1–1.5 m s−1. The climatic conditions of the test room were 22–25°C and RH = 30%. We observed each spider for 10 minutes to record any dispersal behaviour. Sample size was 30 for each treatment.

Activity experiments

In the crab spider, we conducted an additional experiment to test the effect of ant cues on its activity level. We could not perform this experiment with the cobweb spider because its light coloration made it undetectable to the video-tracking software (EthoVision XT 8; Noldus Information Technology, Wageningen, The Netherlands). We established the treatments in a similar way as the previous experiments. We placed moistened white filter papers on the bottom of large Petri dishes with 9-cm diameter and put six Lasius ants, three Formica ants or no ants inside each dish according to the treatment. We enclosed the dishes with fluon-covered tops and we kept them inside a climate cabinet for 24 hours (see Dispersal experiments). Afterwards, we took the filter papers away from the dishes and put them individually into experimental arenas, consisting of one Petri dish at the bottom of a plastic cup that had fluon-coated walls (diameter: 10 cm; height: 15 cm). Then we released one crab spider onto the centre of each experimental Petri dish and filmed its behaviour with a video camera (SONY HDR-CX 550 VE, Sony Corporation, Tokyo, Japan) for 15 minutes. We filmed six arenas simultaneously, randomly assigned to treatments so that we had two replicates of each of the three treatments in all recordings. During the video recordings, we covered the arenas with glass panels to prevent the control arenas from being influenced by potential volatile cues. We used a new Petri dish for each recording.

To quantify the behaviour of the crab spider, we analysed the videos with the software EthoVision XT 8 with the Multiple Arenas Module that tracked the six individuals simultaneously. We recorded 15 videos to reach a sample size of 30 individuals per treatment. We measured the following behavioural variables indicating activity: distance walked, time spent mobile (body movement defined as 5–60% pixel changes per frame of the complete area detected as spider, also while stationary) versus time spent highly mobile (more than 60% body movement) versus time spent immobile (less than 5% body movement); time spent walking (central point of the individual is moving in a directed manner), average speed.

Statistical analyses

To analyse the effect of treatment (cue, control) on spider dispersal propensity, we used generalized linear models with Binomial distribution (logit-link function, corrected for overdispersion) to test for presence/absence of dispersal. We performed four analyses, one for each ant cue–spider species combination.

To analyse the effect of treatment (Lasius cue, Formica cue, control) on the activity of the crab spider, we first conducted a PERMANOVA as an overall test because the behavioural variables were highly correlated. PERMANOVA is a non-parametric analogue to MANOVA that has no requirements on the multivariate distribution of the data. When the PERMANOVA was significant, we tested the behavioural variables individually with general or generalized linear models with the normal or gamma distribution, respectively. The purpose of these subsequent univariate analyses was to detect the direction and magnitude of the change in activity produced by the ant cues. We used R version 2.15.3 (R Core Team, 2013) for all the analyses, with the package ‘vegan’ for the multivariate analyses (Oksanen et al., 2013).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Silk-based dispersal

All spiders that dispersed used the rappelling strategy. The proportion of crab spiders rappelling was more than twice as high when exposed to Lasius21,58 = 6.08, P = 0.014; Fig. 1a) and Formica cues (χ21,58 = 6.91, P = 0.0086) as in the controls. In cobweb spiders, rappelling almost doubled when exposed to Lasius cues (χ21,58 = 4.34, P = 0.037; Fig. 1b), but Formica cues had no significant effect on the rappelling propensity of this species (χ21,58 = 0.63, P = 0.43).

figure

Figure 1. Effect of ant cues (mean ± se) on the dispersal (rappelling) probability of (a) Xysticus and (b) Phylloneta impressa. n = 30 in all experiments (*P < 0.05, **P < 0.01).

Download figure to PowerPoint

Walking activity

PERMANOVA revealed an overall significant effect of ant cues on the walking activity of the crab spider (pseudo-F2,87 = 3.07, P = 0.029; Fig. 2), with Formica cues changing activity levels (pseudo-F1,87 = 5.82, P = 0.008), but not Lasius cues (pseudo-F1,87 = 0.32, P = 0.72). Formica cues significantly increased the time crab spiders spent walking (F1,58 = 7.76, P = 0.0072), the time they spent being highly mobile (F1,58 = 4.91, P = 0.031), the average walking speed (F1,58 = 4.37, P = 0.041) and, marginally, the time they spent being mobile (F1,58 = 3.43, P = 0.069). Formica cues decreased the time crab spiders spent being immobile (F1,58 = 4.23, P = 0.044).

figure

Figure 2. Effect of ant cues (mean ± se) on the walking activity of Xysticus (pseudo-F2,87 = 3.07, P = 0.029), represented by the following variables: distance walked, time spent immobile, mobile and highly mobile; time spent walking, average speed. Because only the overall effect of Formica cues was statistically significant (pseudo-F1,87 = 5.82, P = 0.008), we did not include the Lasius cue treatment in the univariate tests of behavioural variables. n = 30.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

We found that ant cues induced spiders to disperse. The reduced response to ant cues by the web-building cobweb spider compared with the free hunting crab spider suggests that the use of shelters by spiders may modulate the risk posed by ants. Our experiments suggest spider dispersal in response to ant cues as a mechanism underlying the strong negative effects of ants on spider densities in the field and underscore the importance of behaviour in our understanding of species interactions.

More than twice as many crab spiders dispersed in the ant cue treatments than in the control treatments, the effect of Formica cues being stronger than Lasius cues. Interestingly, the video analyses revealed an increase in walking activity when spiders where exposed to Formica but not Lasius cues. Thus, both walking and rappelling were more strongly affected by Formica than by Lasius. Compared with Lasius, Formica act more often as predators and include a higher fraction of arthropods in their diet (Seifert, 2007). Thus, it can be assumed that the presence of Formica poses a higher risk to spiders, which can explain the stronger effects of Formica cues on the behaviour of the crab spider. By contrast, Formica cues had no effect in the dispersal propensity of the cobweb spider, whereas Lasius cues increased it but to a lesser extent than in the crab spider. Unlike crab spiders (Thomisidae), cobweb spiders (Theridiidae) build three-dimensional tangle webs that have a shelter among the threads. This kind of webs and shelters can protect spiders from predators such as ants and wasps (Blackledge, Coddington & Gillespie, 2003; Garcia & Styrsky, 2013; but see Coudrain, Herzog & Entling, 2013), so the degree of risk posed by ants patrolling among the vegetation may be lower in cobweb spiders compared with the free-hunting crab spiders. It is not obvious why this protection should be more effective against Formica than Lasius. Thus, the differentiation of Phylloneta between Lasius and Formica may have other reasons. For example, Lasius is more abundant than Formica in grasslands, a preferred habitat by Phylloneta (pers. obs.; Hänggi, Stöckli & Nentwig, 1995; Seifert, 2007), which makes Lasius the most relevant interacting species. Because the presence of Lasius reduces densities of many types of insects (Piñol et al., 2012a), Phylloneta could be avoiding Lasius cues as an indication of decreased prey availability.

Many field experiments found that ants negatively affect insect and spider densities, but did not investigate the mechanisms involved in these interactions (Vandermeer et al., 2002; Sanders et al., 2011; Piñol et al., 2012a,b; Mestre et al., 2013). Mestre et al. (2013) excluded ants from tree canopies for 8 years and analysed the changes in the spider assemblage. The study revealed that ants (mainly Lasius) especially reduced the densities of spider species with a sedentary lifestyle, Xysticus spp. and many Theridiidae species among them. In the present study, we purposely tested two spider species closely related to those most negatively affected by the presence of patrolling ants in Mestre et al. (2013) and found that spiders increased their propensity to emigrate when confronted to cues of ant presence. Thus, our laboratory experiments show that ant TME on spider dispersal may be one of the mechanisms behind the ant-induced reduction in spider densities reported in different field studies. A further research question would be the relative importance on spiders of ant TME relative to ant predation, while bearing in mind that in natural communities ants can also exert an influence on spiders through indirect effects (i.e. effects that go beyond adjacent trophic levels, such as exploitative competition; Morin, 2011). Moreover, only a handful of studies have examined the occurrence of ant TME on arthropods: laboratory bioassays show that ant cues affect the oviposition behaviour of flies and beetles (Oliver et al., 2008; Van Mele et al., 2009) and field experiments prove that ant presence disrupts foraging behaviour in folivores although they do not test the role of chemical cues on these behavioural changes (Messina, 1981; Rudgers, Hodgen & White, 2003). Thus, it remains to be seen whether the effect of ant cues found in the laboratory is also detectable in the field. A drawback in our laboratory experiment is the way we had to handle ants to introduce them into the arenas (see ‘dispersal experiments’ in the Methods section). These manipulations with the pooter most probably stressed the ants and made them release alarm pheromones. Thus, it is possible that alarm pheromones introduced a confounding effect in the experiment, because we were interested in testing the effect of cues left by patrolling but not by stressed ants. The cues deposited on the filter paper likely contained trail pheromones and above all cuticular hydrocarbons, which are secreted by the feet and are passively deposited while walking (Lenoir et al., 2009). Nonetheless, cuticular hydrocarbons are non-volatile and long-lasting (Martin, Helanterä & Drijfhout, 2008; Lenoir et al., 2009), whereas alarm pheromones are volatile (Löfqvist, 1976; Attygalle et al., 1987) and ants were not stressed when we opened the arenas, so alarm pheromones probably evaporated before we placed the spiders on the arenas, although we cannot completely rule out this confounding effect.

Our study concentrated on the effect of ant cues on spider dispersal behaviour. Silk-based dispersal in spiders is costly in terms of both energy and risk, because silk production is energetically demanding, and spiders, as passive dispersers, cannot control the dispersal path (Bonte et al., 2012). Hence, the fact that ant cues increased spider dispersal indicates that the costs of being exposed to patrolling ants or to reduced prey densities are higher than the costs of emigrating. Moreover, we found that ant specific identity caused differences in dispersal propensity, which suggests that spiders modify their emigration decisions in accordance to the degree of perceived risk in the current location. Silk-based emigration takes place in response to suboptimal environmental conditions and spiders can cue on different factors as a source of environmental information, such as prey availability, maternal effects, degree of inbreeding and conspecific density (De Meester & Bonte, 2010; Entling, Stämpfli & Ovaskainen, 2011; Mestre & Lubin, 2011; Mestre & Bonte, 2012). Our study now shows that cues of predation risk also play a role in spider dispersal and it is one of the few that experimentally demonstrates the influence of TME on dispersal behaviour in terrestrial systems (Moran & Hurd, 1994; Weisser, Braendle & Minoretti, 1999; Mondor, Rosenheim & Addicott, 2005). Natural communities are embedded in a spatial matrix interconnected by migrating individuals. Despite the crucial importance of dispersal in metacommunity dynamics, much progress has yet to be made in quantifying dispersal traits of species and in characterizing the spatial properties of species interactions (Leibold et al., 2004; Massol et al., 2011).

In conclusion, spiders are able to perceive ant cues and adjust their dispersal decisions to the degree of risk. This underlines the importance of TMEs in intraguild interactions between terrestrial predators. Moreover, our study suggests that TMEs contribute to the impact of ants on arthropod communities.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

We are grateful to Franziska Möller for the help with spider rearing and to Florian Menzel for the advice in keeping ant colonies. This work was supported by the Ministerio de Ciencia y Tecnología-Fondo Europeo de Desarrollo Regional (MCYT-FEDER; grant numbers CGL2007–64080-C02-01/BOS and CGL2010–18182), by the Ministerio de Ciencia e Innovación-Fondo Social Europeo (MICINN-FSE; fellowship number BES-2008–007494) and by the Deutsche Forschungsgemeinschaft (DFG; grant number EN979/1-1). Arthropods were collected under the permit 42/553-233 issued by the Struktur- und Genehmigungsdirektion Süd, Neustadt an der Weinstrasse.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References